Twenty-four jars of jam on a folding table draw a crowd; six jars draw fewer browsers. The table with more jars also converts far better — about ten times the purchases. That is the jam study, Iyengar and Lepper (2000, Journal of Personality and Social Psychology 79(6), n=754): 24 varieties against 6, roughly 60% of passersby stopping against 40%, and 3% buying against 30%. The two conditions alternated across Saturdays in a single supermarket, which is a narrow stage for so famous a result. Yet the number — a tenfold swing in purchases from the mere count of options — is the one most often pulled out to argue that abundance paralyzes.

Here is the part the slide decks skip. Scheibehenne, Greifeneder and Todd (2010) pooled 50 experiments with 5,036 participants and found the average choice-overload effect to be, in their words, virtually zero; a related restatement puts 63 studies at d≈0.02. Chernev, Böckenholt and Goodman (2015) went further with 99 observations and 7,202 participants, finding no overall overload at all — only conditional effects that surfaced when sets were complex, tasks difficult, preferences uncertain, or goals vague. A 2024 replication across seven scenarios found too much choice in about 35% of cases and the opposite effect, more choice helping, in about 20%. The honest reading is not that the jam study was faked. It is that the effect is near zero on average and highly conditional, which is a very different thing from a law of product design.

Spread the anchors side by side and the conditionality becomes visible:

AnchorSource and yearWhat was comparedReported resultWhat it does not establish
Jam counterIyengar & Lepper 2000, JPSP 79(6)24 versus 6 varieties, one supermarketStopping 60% versus 40%; buying 3% versus 30%A tenfold effect that holds in every setting
Broad poolScheibehenne et al. 2010, 50 experiments, N=5,036Many versus few options across studiesAverage effect virtually zero (63 studies: d≈0.02)That overload never occurs
Conditional poolChernev et al. 2015, 99 observations, N=7,202Complex versus simple sets, hard versus easy tasksNo overall overload; depends on four conditionsA fixed penalty for any large set
Replication2024, seven scenariosOversized set versus a lean setOverload in about 35%, reverse effect in about 20%One stable direction of effect
Storefront growthAppMagic 2026 report; AppfiguresLaunches versus new submissions, 20251.4M+ launched, 5.4M+ across two stores; ~68% under 1,000 downloadsA single agreed count of catalog size

The comfortable slogan — that fewer options always convert better — is not what the evidence supports. Cutting options at random is just as likely to remove the very thing that let someone recognize a good fit. What the conditional work points to is structure rather than scarcity: a large set that is grouped, staged and labeled behaves differently from the same set dumped flat. Nielsen Norman Group's long-running guidance lands in the same place. More options mean longer decisions, and the recommended remedies are grouping, chunking and progressive disclosure, not blunt deletion. The aim is not to shrink a catalog but to make its shape legible. An unorganized wall of nine near-identical entries and a clean wall of nine near-identical entries are different products even though the count is identical.

The commercial data, read carefully, agree. Apple's own figures say about 70% of visitors find apps through search and 90% never scroll past the tenth result; discoverability, not sheer count, decides what gets seen. Long and colleagues (2025, M&SOM) found that larger recommendation sets lowered purchase probability, with up to 64% of the drop coming from visitors who clicked nothing at all — a signature of hesitation rather than an active rejection. One retailer that trimmed categories reported 7% more category spending and 5% faster shopping, though that is vendor-reported and directional rather than a controlled estimate. The storefront numbers need the same care: AppMagic's 2026 landscape reports 1.4M+ apps launched globally in 2025, up about 25%, with around 10% gaining real traction and roughly 68% never reaching 1,000 downloads; the two major stores together hold 5.4M+ apps. Appfigures puts 2025 new submissions at 557,000 — submissions, not launches, a different measure that cannot be swapped in. None of this proves that fewer options cause more sales. These are correlations and self-reports pointing one way: unorganized abundance carries a cost, and the cost shows up first as a visitor who never clicks.

A team running an open catalog has a more useful move available than rationing programs. Give the directory a shape: group by task and capability, reveal detail in stages, and make the first screen a map rather than a wall. On a shared runtime that structure can be versioned and inspected instead of being baked into each listing, so a grouped view can be tested against a flat one and the winner kept on evidence — specifically, on how many visitors leave without clicking anything. Cross Mini App keeps those catalog surfaces in the open directory so the comparison is reproducible by anyone auditing it, not only by the team that shipped the layout. Nine options are not the problem. Nine options with no order are.

Sources